PQEventCog: Classification of power quality disturbances based on optimized S-transform and CNNs with noisy labeled datasets

IIR Shelving Filter, Support Vector Machine and k-Nearest Neighbors Algorithm Application for Voltage Transients and Short-Duration RMS Variations Analysis

“…It must be highlighted that many authors boast about the very high classification accuracy achieved by their algorithms: for example, under different noisy environments, the declared classification accuracy is 99.4-100% in [17], 93.3-100% in [20], etc. Nevertheless, neither these nor other PQ classifiers are currently used in practice.…”

“…In our opinion, in anticipation of a great leap forward in the PQ field, authors should be more honest (transparent) and discuss the limitations of their methods and approaches (e.g., as seen in Section 4.1). The authors sometimes also highlight further development avenues: for example, [20] mentions light-weighted optimization required for deploying the algorithm in embedded systems and also online training of its PQ classification model. A great diversity of feature extraction approaches can also be noticed in Table 1, and all currently investigated techniques has a potential for wide usage in the future.…”

“…19 The following parameters are required for voltage sag generation with the synthetic signal generator: signal duration, fault start time, fault duration, nominal RMS, signal frequency, sampling frequency, no-fault amplitude, phase angle offset, number of smoothing timepoints, fault severity, signal rotation, noise percentage, and the label (the type of the fault to be generated). 20 The ensemble utilizes discriminant, KNN and DT weak learners (three learners, nine methods). 21 Three causes are included: fault, motor inrush current, and transformer inrush current.…”